Research And Application Of Image Registration In Panorama Stitching

Panorama is a kind of wide angle diagram which exists in. the form of paintings, photos, video and3D models. Modern panorama is difined as photos token by camera and processed in computer. Mosaicing images with some coincidence token by a fixed camera by hand is an commen method to achieve panorama. As manual method needs strict shooting conditions, the automatic panoramic image mosaic technology has been developing. In order to realize the panorama stitching, we need to find a space transformation relation which maps an image to another through corresponding points in different images representing the same object.In this paper we analyze the whole process of image stitching and make improvement on the traditional algorithm aiming at the application of panoramic mosaic. The key improvement are:(1)In the image pre-processing stage, we capture some parameters of mobile sensor while taking picture. Through the rational use of these parameters we can simplify the image registration procedure. First, We can calculate the focal length of each image by these parameters to zoom images to the same scale. Second, we can achieve a more simple cylindrical transform by the limits of moblie phone sensor. Third, we obtain a simple image registration method by means of sensor parameters.(2) In the image registration stage, we propose a new feature point based on Harris and SIFT algorithm by which we complete image registration and guarantees the accuracy with higher efficiency. This paper analyzes several matching strategies to improve the matching accuracy. Then, we use RANSAC algorithm combining with the least square method to calculate the transformation matrix and get an accurate result.(3) In the image fusion stage, this paper realizes the traditional nibbling fusion method and proposes a new fusion method by normalizing feature points’ pixel value to get a natural panorama.Through the simulation of the algorithm proposed in this paper, we get a good result. Aiming at the scene of panorama stitching, the method we proposed is more suitable than the traditional method. Finally, we developed an application of panorama on mobile phone.